What the end of third-party cookies means for personalization

This article is part of a VB special issue. Read the full series here: How Data Privacy Is Transforming Marketing.

We’ve been shaking the crystal ball on the cookieless future, and it’s still cloudy — we know it’s coming, but we’re not sure when, or how exactly it will play out.

Still, now is the time for organizations to prepare, lest their marketing methods become obsolete.

It is imperative, experts say, that enterprises be proactive in balancing the dual consumer demand for privacy and personalization. How can they achieve this? By harnessing lower-level types of data — including second-party, first-party and zero-party — and leveraging artificial intelligence (AI) in a way that is both ethical and accurate. 

“Moving forward, brands have to think about how to collect data transparently and use it in a way that delivers value to the customer,” said Stephanie Liu, privacy and marketing analyst at Forrester. “That’s a relatively new mindset for marketers, and many are struggling today because for decades they’ve prioritized benefits to the business while neglecting the customer.”


Low-Code/No-Code Summit

Join today’s leading executives at the Low-Code/No-Code Summit virtually on November 9. Register for your free pass today.

Register Here

Comparing first-party data and third-party data

Essentially, first-party data is “data that customers and companies share ownership of,” said Andrew Frank, VP analyst at Gartner. This lets a brand tailor experiences in the way of loyalty programs and incentives. 

Putting it in human terms: First-party data is like being friends with someone and sharing information directly, said Liu. 

“You know each other well and your friendship can deepen over time,” she said. 

Third-party data, by contrast, is akin to having an acquaintance who you’ve mostly heard things about “through the grapevine” — and not all that is accurate.

“Personalization has turned into an amorphous catch-all, but when it comes to asking customers for data, brands need to think about what data they need, how they’ll use it to benefit the customer and how they’ll encourage customers to actually share that data,” Liu said. 

With changes occurring and more afoot, “marketers are facing data deprecation,” she added.

Cross-site tracking is becoming more difficult, privacy regulations are adding new consent requirements, consumers are more protective of their data and walled gardens are limiting data access and use. 

“It’s not just the death of third-party cookies,” said Liu. “There are multiple significant forces impacting marketers’ ability to collect and use customer data.”

The power of AI

Organizations are increasingly leveraging AI to fill in this gap. AI and machine learning (ML) models can categorize and segment third-party data to correlate, segment and make predictions. 

Liu pointed to one common use case of lookalike modeling. When a customer hasn’t shared “a plethora of information about themselves,” a brand can take what it does know about them and try to match them with customers who look similar, she explained. 

“It’s a way of filling in the gaps for customers whose profiles are data scarce,” said Liu. 

Unsurprisingly, there are risks. If someone has chosen not to share much about themselves, it’s probably because they don’t know the brand well or don’t see value in sharing data, she pointed out. 

Brands can nail it pretty accurately and personalize based on data a customer hasn’t explicitly shared, but this can be perceived as “creepy and invasive,” she said. Case in point: The infamous example of Target recognizing a customer was pregnant before she’d even broken the news to her own father. 

On the other hand, if a brand gets it wrong, it risks personalizing off faulty assumptions. 

“So, marketers need to think about what benefit the customer will get from this type of modeling and if the benefits (to marketers) are worth the risks (to customers),” said Liu.

‘Small data’ trend

The conventional wisdom is that the most cutting-edge AI is dependent on large volumes of data. However, other approaches do not require massive labeled datasets — a few examples are transfer learning, data labeling, artificial data, reinforcement learning and Bayesian methods, according to the Center for Security and Emerging Technology. 

This is what’s known as “small data.”

“Behaviors have changed so much in so many different ways in society around the world, that the data you collect is less indicative of the future than it used to be,” said Erick Brethenoux, VP analyst at Gartner. 

Organizations may have a lot in terms of volume, but not quality, he said. And, when there isn’t enough quality data or data is fragmented, that’s when model accuracy decreases.

This is prompting the use of additional AI techniques in the background to “enhance or complement” data, said Brethenoux. For example, in insurance, applying knowledge graphs to provide more context and better accuracy. 

“The people who say they have too much data don’t know what is in their data,” said Brethenoux. 

Other types of data collection

But, as third-party data from cookies to fuel AI models decreases, brands can increasingly rely on another tool: “Zero-party data.” 

This was termed by Forrester in 2017, and it refers to data that a customer proactively and intentionally volunteers. Such as, Liu said, product preferences, purchase intent, and content preferences. 

For example, they can specify, “I have a cat.” A brand can then use this information to show them cat products on their site or app — and stay away from hawking dog products. 

“This is data customers are choosing to share with a brand because they like the brand and are getting some benefit or value in return,” said Liu. It is much more transparent and straightforward than buying from a data broker, she said, and helps reduce the creepiness factor of “why do you know that about me?”

Right now, it’s still just a concept, contended Frank. He does see it evolving into something “more substantial,” and potentially used with distributed or decentralized ledgers. 

Still, he pointed out that first and zero-party data, where there is “incentivized consent” is not always permitted — or even a possibility. More generic categories that don’t sell directly — say, a tissue paper supplier — don’t have that ability, and the cost of losing access to third-party data is higher.

Second-party data

Another emerging method for procuring data? Second-party data via data clean rooms. This is a collaboration between brands with direct relationships to consumers with brands that don’t, explained Frank.

Data clean rooms allow companies to leverage intelligence extracted from personal data without exposing personal data to any parties, he explained. 

A new Interactive Advertising Bureau standard is “seller defined audiences,” which allows companies with large amounts of data to define an audience that an advertiser could buy without revealing specifics, he said.

Then there are concepts such as Unified ID 2.0, an unencrypted alphanumeric identifier created from emails or phone numbers. This method allows advertisers to target specific consumers without compromising their privacy.

Responsible AI — and marketing

The key to all this is getting the right kind of consent, and making sure that that is always honored and enforced in different contexts.  

Then, of course, there’s the imperative that AI models be responsible, ethical and trustworthy — undoubtedly one of the most pressing discussions occurring in tech right now. Respective to third-party data, organizations must be cautious and seek advice on how to use it, said Brethenoux.

“It is the responsibility of the organizations getting that data to do that work,” he said.

The future of procuring data, said Frank, could either be a “walled garden” concept, where a few large companies have a great wealth of data and sell that data; a “consent economy” controlled by consumers; or a decentralized, self-sovereign identity where people would control their identity.

In any case, “we’re heading for a world where people do have more control over their personal data and can make more intelligent decisions with how they share it with brands,” said Frank. 

Ultimately, third-party data isn’t going to go away, said Frank. Brands just must get smarter about how they use all other types of available data — whether that’s zero, first, second, or creative procurement of third-party data that respects privacy. 

In the meantime, continue to keep an eye on that cookieless crystal ball. 

Originally appeared on: TheSpuzz